Abstract

When working in agricultural environments, specially in groves with dense foliage, machinery positioning systems might suffer from loss of GNSS (Global Navigation Satellite System) signal. The latter motivated the development of new localization strategies that use the environment information to localize the machinery and thus fulfil the required agricultural task. In this work, the usability of five well known scan matching algorithms as sole localization systems using a 2D LiDAR (Light Detection and Ranging) scanner is tested in an avocado grove. The aim is to show the pros and cons of such techniques when the machinery faces a real agricultural environment: presence of slippage, absence of GNSS signal, non-flat terrains in a non-experimental grove and noisy LiDAR readings. The analysis presented herein concludes with a localization error evaluation when the machinery has to travel through a rough avocado alley, showing that amongst all the techniques implemented, the Probabilistic Iterative Correspondence (PIC) and the Sum of Gaussian Scan Correlation (SGSC) presented the lowest localization estimation error and remained consistent from a localization point of view.